3dfx SLI comeback!

[3dfxFanatic]

NVidia has begun usin sli in the 6800 series
read here:
http://guru3d.com/newsitem.php?id=1628

reminds me the good old days when voodoo2 owned the world...

[lecram25]

*SIGH*

nVidia's SLi != 3dfx's SLI

I will say this for the last time.

3dfx's SLI (if you could even call it 3dfx's) belongs to an individual; that person holds analog SLI whilst 3dfx held/owned digital SLI. (this person will remain nameless unless he/she feels that it is necessary for everyone to know who he/she is)

This new stuff that nVidia has brought out is a form of digital SLI. But it's not even SLI in it's right form. nVidia couldn't even use the right words for the "SLI" acronym. Instead, they were forced to call it "Scalable Link Interface" as to avoid patent infringement.

nVidia very well knows that by them assigning their "new" technology the acronym "SLI", would make everyone go crazy over it, figures.

Guess what? It's called MARKETING, and believe it or not, it's disgustingly working.

[Voodoo5]

If 3dfx had the rights to use SLI (scan line interleave)would'nt this technology have been transfered to Nvidia when they acquired 3dfx ?

Why also is it so secretive as to who owns the rights SLI ?

Just curious

[lecram25]

There are two versions of SLI, like I said before: digital and analog.

digital was 3dfx's, and analog belongs to an individual. m'kay?

[amp_man]

This new SLi is entirely different, instead of rotating every other line of a texture, like 3dfx used (at least in analog, not sure about digital), Nvidia's method has one card drawing the top half of the screen, and the other card draws the lower half, which, IMOHO, makes very little sense, because one card could wind up doing almost all sky, whereas the other card has everything on the ground. Of course, the drivers do some load balancing, so it isn't exactly "half and half", but still it doesn't seem as efficient as the so-called "3dfx method." But this is Nvidia, so they can likely pull it off somehow, and if they do, methinks they now own this chapter in the graphics card wars. ATi has something in the works (tile based? can't remember), but I doubt they'll have it working in time to compete.
OTOH, this now means that you'll need 4 molex connectors, 2 x16 PCI-E connectors, and a motherboard and CPU to match in order to run such a configuration, which is about a 400W power drain before you even get to HDDs, etc. A very costly powerhouse, if you ask me (lecram, did you see that mastercard spin-off? I can't seem to find it again...)

EDIT: wow, my spelling sux this late at night

[lecram25]

This new SLi is entirely different, instead of rotating every other line of a texture, like 3dfx used (at least in analog, not sure about digital),

You have no idea how SLI works.

OTOH, this now means that you'll need 4 molex connectors, 2 x16 PCI-E connectors,

Wrong. The new TSMC only require one molex connector from what my sources are telling me.

(lecram, did you see that mastercard spin-off? I can't seem to find it again...)

[3dfxFanatic]

Originally posted by amp_man



This new SLi is entirely different, instead of rotating every other line of a texture, like 3dfx used (at least in analog, not sure about digital), Nvidia's method has one card drawing the top half of the screen, and the other card draws the lower half, which, IMOHO, makes very little sense, because one card could wind up doing almost all sky, whereas the other card has everything on the ground. Of course, the drivers do some load balancing, so it isn't exactly "half and half", but still it doesn't seem as efficient as the so-called "3dfx method." But this is Nvidia, so they can likely pull it off somehow, and if they do, methinks they now own this chapter in the graphics card wars

did you think nvidia is that stupid?!
they still have some good 3dfx engineers there...
the new method work like you said, but, before each card is given the half scene to render, a calculation is made, and if one half needs much more rendering power, the card with the other half gets to render a bigger part of the scene, so it makes it much more efficient... and performance wise

[amp_man]

Wow, Let the flaming begin!

Originally posted by 3dfxFanatic
did you think nvidia is that stupid?!

do I really have to answer that?

they still have some good 3dfx engineers there...

and what does that have to do with anything??? Just because somebody worked for 3dfx 3 1/2 years ago, they're automatically given genius status?!?!

the new method work like you said, but, before each card is given the half scene to render, a calculation is made, and if one half needs much more rendering power, the card with the other half gets to render a bigger part of the scene, so it makes it much more efficient... and performance wise

Yes, I'm aware of this, I've probably read just as many articles about it as you have. but how complex are those calculations going to be? IE how much of a performance hit are such drivers going to make on the CPU? Anyways, I guess they are getting some efficiency out of it, close to 77% from what I've heard (but then again, you shouldn't trust everything you hear, especially on the net...8 real, 8 fake pipelines, etc.)

You have no idea how SLI works.

And I could care less, honestly. It's old tech, and it works, atm, so why bother researching up on it? I know how analog SLI works, but I really haven't cared enough to research digital. And what relevence does that have?

Wrong. The new TSMC only require one molex connector from what my sources are telling me.

Okay then...again, what relevence? Ooh, only two molex connectors, still gonna need a ton of power.

Crap, can't find the mastercard thing...oh well, later boys

[lecram25]

When someone says it'll down to one molex over two clearly means that there had to be a power readjustment, in the positive direction.

No, you don't know about SLI. You talk about "rotating every other line of a texture".

Must have been that late night typing.

[amp_man]

Originally posted by lecram25

When someone says it'll down to one molex over two clearly means that there had to be a power readjustment, in the positive direction.

True, there has been a small decrease in the amount of power used by the card...but not much! Check the specs, I'm too lazy to.

No, you don't know about SLI. You talk about "rotating every other line of a texture".
Must have been that late night typing.

Yes, very late, about 1 am or so, I'm guessing, and it's about midnight right now. Sorry, I didn't exactly mean "rotating", but more like "switching back and forth" or something. Each card renders every other line of a texture, which is then combined in the primary card and sent to the monitor. Any better?

[ArchAngelCD]

My only opinion on this subject is....
LONG LIVE ATI!!